讨论了不确定性规则的知识系统同神经元网络相结合的问题;提出了适合处理不确定性规则的负梯度学习算法,使不确定性规则能方便地转换成神经元网络的权矩阵;研究了如何利用神经元网络的生成能力来获取新的不确定性规则,使知识系统具有知识自动获取能力,并使原有规则系统的脆弱性得以改善. The problem of combining the knowledge system of uncertain rules with neural networks is discussed. A negative gradient learning algorithm suitable for dealing with uncertain rules is proposed so that the uncertainty rules can be conveniently transformed into the weight matrix of neural networks. How to use the generation ability of neural network to obtain new rules of uncertainty, make the knowledge system acquire knowledge automatically and improve the vulnerability of the original rules system.